Dwarkesh Patel's 8 predictions nail how continual learning will torch today's static LLMs, birthing AIs that ingest fresh data nonstop and evolve without resets. This shift matters because it accelerates real autonomy—think agents rewriting their own weights from live interactions, not retraining cycles. My take? He nails the upside on personalized intelligence but glosses the chaos: unchecked adaptation risks runaway misalignment faster than any lab can patch. Static models were a crutch; this is the match. Which of his calls do you see blowing up first? Full video link https://youtu.be/iewm45atodE